Re: st: Dirifit / ddirifit: "maximum number of iterations exceeded"

[Maarten buis <maartenbuis@yahoo.co.uk>]

--- On Sun, 25/4/10, Martin Gelter wrote:
> Depending on which control variables we use, we sometimes
> do not get results at all (dirifit reports non-converging
> iterations until the user stops it).

This usually indicates that there may be something like near
perfect colinearity, or badly scaled covariates (extremely
huge number or extremely small numbers), or near perfect 
predictions. I would start with rescaling my explantory
variables. You can also try -fmlogit-. (if you care about
the technical details: -fmlogit- is implemented as -d2-, that 
is with analytical first and second derivatives, which means 
that it is quicker, but also that it can handle more difficult
estimation problems)

> Sometimes, when we do get a result with dirifit, ddirifit
> will result in an error message saying "maximum number of
> iterations exceeded" (I typically set the maximum number up
> to 10000). We suspect that one of the problems could be that
> we have many values in the matrix that are very close to
> zero.

The newest version of -ddirifit- should no longer give you
this error message, as it now calculates the delta method 
standard errors analytically. 

Hope this helps,
Maarten

--------------------------
Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen
Germany

http://www.maartenbuis.nl
--------------------------


      

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/